公开(公告)号：US20160103758A1

公开(公告)日：2016-04-14

申请号：US14509741

申请日：2014-10-08

Applicant: Yahoo! Inc.

Inventor： Zhenyu Zhao ,  Flavio T.P. Oliveira ,  Maria Stone ,  Miao Chen ,  Shalu Pandey ,  Kshitiz Tripathi

IPC: G06F11/36 ,  G06F9/445

CPC classification number: G06F8/65 ,  G06F11/3684 ,  G06F11/3692

Abstract: The technologies described herein use a statistical test to determine whether differences between data sets of buckets in a bucket test, such as differences between averages of two buckets (e.g., differences between means of two buckets), are directionally larger than a predetermined or preset minimum threshold value. The statistical test may also provide an extension to specify the minimum threshold value as a percentage. Also, described herein are techniques for estimating different control variables of a bucket test, such as estimating minimum bucket size to provide sufficient statistical power with use of the minimum threshold value.

Abstract translation: 本文描述的技术使用统计测试来确定桶测试中的桶的数据组之间的差异，例如两个桶的平均值之间的差异（例如，两个桶的装置之间的差异）是否定向地大于预定的或预设的最小值 阈值。 统计测试还可以提供一个扩展，以百分比形式指定最小阈值。 此外，这里描述的是用于估计桶测试的不同控制变量的技术，例如使用最小阈值来估计最小桶大小以提供足够的统计功率。

公开(公告)号：US20160364745A1

公开(公告)日：2016-12-15

申请号：US14734515

申请日：2015-06-09

Applicant: Yahoo! Inc.

Inventor： Yan He ,  Miao Chen ,  Maria Stone

IPC: G06Q30/02

Abstract: Online experimentation has been widely used to evaluate an effect of a new feature of an online product on user engagement. One challenge is that an existence of outliers can often complicate the analysis of such experimental results. Thus, a procedure is provided herein to detect and remove outliers from experimental results. The procedure can use statistical tests based on parametric distributions of sample maximum or minimum. These tests can be performed using an inward testing procedure to identify multiple outliers. Finally, these filtered test results can be used to control delivery of a new feature of an online product.

Abstract translation: 在线实验被广泛用于评估在线产品的新功能对用户参与的影响。 一个挑战是，异常值的存在往往会使这些实验结果的分析复杂化。 因此，本文提供了一种从实验结果中检测和去除异常值的过程。 该过程可以使用基于样本最大值或最小值的参数分布的统计检验。 这些测试可以使用向内测试程序进行，以识别多个异常值。 最后，这些过滤的测试结果可用于控制在线产品的新功能的传递。